Dear Francesco,
I thank you very much for your answer! For my purposes I guess two-pass
encoding remains the best choice, but I would like to find the time to
make some practice with this option, at least for curiosity.
Best regards.
Francesco Romani wrote:
On 2/5/07, *Andrea Giuliano* <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
Hi,
I'm here again with a question about quality and output file size.
Looking at the man pages I noticed that the value "3" for the "-R"
option (aka "constant quantizer encoding") is documented quite less
than
the other values. What sounds quite strange to me is the total absence
of a comparison with the other modes.
Before I start a lot of experiments with this options, can someone
direct me to a deeper documentation, or just give me a hint?
In a nutshell,
``quantizers'' are one of major sources of compression in MPEG-like
algorythtms: using those values the encoder internally tune the
compression ratio
of a given sequence. You can see those (VERY loosely) as a compression
index:
more they are low (near to zero), more the encoder is able to save
details in compressed
picture. Of course this doesn't come for free. A more detailed pictually
usually needs
more information, so more bits to be encoded.
During (lossy) compression, there always is a tradeoff between space
(used to represent
informatio to be encoded) and encoding quality. Given the algorythm, a
better image
always require more bits. You can fix or the output quality using a
constant quantizer,
but doing so you lose control about final output size, or you can fix
the final output size,
but you lose control about image quality.
Of course `lose' is a too strong term since user always has some options
to tune to improve
things.
+++
Now in italian, same as above, rephrased (and hopefully made clearer :P)
Molto in breve,
I `quantizer' sono uno degli aspetti principali con cui i codec
MPEG-like ottengono la compressione
(con perdita). Vengono usati principalmente per controllare la qualita`
della compressione influenzando
la quantita` di dettagli che viene riprodotta nell'immagine codificata.
Ad esempio, con codec MPEG4
tipo XviD o ffmpeg, immagini codificate con quantizer alti tendono ad
essere blocchettose.
Chiaramente c'e` sempre un compromesso tra qualita` dell'immagine e
dimensione del prodotto
codificato finale.
Se si fissa la dimensione dell'output (cosa che si fa solitamente con la
codifica ad uno o due passi),
l'encoder deve sacrificare la qualita` dell'output per far stare tutto
nello spazio allocato; viceversa,
se si fissa la qualita` (ovvero il livello di dettaglio delle immagini,
ovvero i quantizer da adottare),
allora si deve sacrificare spazio: il prodotto finale verra`
verosimilmente piu` grosso.
Saluti,
--
Francesco Romani
--
Andrea
